PSI - Issue 84

Roberto Acerbis et al. / Procedia Structural Integrity 84 (2026) 765–772

768

The monitoring layout specifies the number and positions of installed sensors and is accompanied by an explanatory note describing the measured variables in relation to the identified structural, geotechnical, hydraulic, and landslide related risks. 2.3. Plant Engineering definition Based on the sensor types and positions defined in the monitoring layout, the monitoring system engineering design is developed, focusing on system architecture (EDGE device locations, cable routing, and configuration of connections between sensors and local acquisition units). A maintenance procedure for the monitoring system is also defined. 2.4. IOT Platform for Data Collection Data from the monitoring systems are collected in a single centralized platform, ARGO IoT, which allows data visualization and download. For each structure, a simplified BIM model is available and integrated with the sensors installed on structural elements. The operational status of each monitoring system is available in real time. 2.5.1.1. As-Built Documentation Verification Before data analysis, it is necessary to verify the correspondence between the designed monitoring system and the system actually implemented. This verification includes comparison between design and as-built layouts, validation of sensor nomenclature and installation coordinates, and review of commissioning and calibration documentation. Photographic documentation further supports visual verification of correct sensor installation and identification of potential installation anomalies. Preliminary data validation is essential to ensure the reliability of subsequent analyses. An anomaly is defined as any condition in which acquired data are not coherent, stable, or reliable for engineering interpretation due to instrumental malfunctions, transmission errors, or incorrect configurations. The validation process is organized into hierarchical checks. First-level checks verify data availability for all sensors and compliance with the expected sampling frequencies (1 sample/min for static quantities and 100 Hz for dynamic quantities). Significant temporal data gaps represent an initial indicator of acquisition system issues. Subsequently, statistical analyses of time series are performed to assess signal stability and identify anomalous values. Parameters such as standard deviation, skewness, and kurtosis, computed on moving averages with appropriate window sizes, allow detection of sensor drifts or inconsistent behavior. Additional checks address physical consistency by verifying the plausibility of measured values and coherence among sensors of the same type installed in similar positions. For static sensors, correlation analysis between measured quantities and structural temperature plays a central role. Linear regression allows quantification of thermal sensitivity and assessment of signal quality through residual analysis and confidence intervals. Significant deviations from expected behavior indicate potential structural or instrumental anomalies. For dynamic sensors, validation includes frequency-domain analysis of accelerometric signals, performed through filtering within expected structural frequency ranges and power spectral density estimation using the Welch method. Analyses are conducted on records acquired under different traffic conditions to distinguish structural response from instrumental noise. 2.5.1.2. Data Pre-Processing Following validation, a pre-processing phase is required to ensure signal consistency, comparability, and direct usability in subsequent engineering analyses. This phase includes several complementary operations. 2.5. Data Analysis 2.5.1. Data Quality Criteria and Validation